Load Forecasting Initiative
Load forecasting is key for many grid decisions across operational and planning timescales. This means that improving the forecasts can lead to improved outcomes, such as more efficient investment decisions and grid performance. However, load forecasting is becoming more complicated due to drivers such as electrification, extreme weather, changing customer behaviors and more. At the same time, these drivers are contributing to increased forecast uncertainty. There is therefore a need to evolve load forecasting and address the critical needs that exist across the industry.
The following three Load Forecasting Initiative workstreams aim to address these critical needs:
- Industry Coordination
- Long-Term Forecasting for Planning
- Short-Term Forecasting for Operations
AI Forecasting for Operations During Extreme Events
This report demonstrates the application of artificial intelligence weather prediction (AIWP) models for power system operations during extreme events, specifically, the use of AIWP models to generate large ensemble forecasts to inform operators of high-risk periods. We use GenCast—an open-source AIWP model from Google—to illustrate this concept through two historical extreme events: Winter Storm Elliot in December 2022 and the Los Angeles wildfires in January 2025. In both cases, the AIWP ensemble forecasts successfully capture the large-scale temperature and wind anomalies several days in advance, providing probabilistic insights into timing and intensity. Furthermore, we demonstrate leveraging the ensemble weather forecasts to produce ensemble load forecasts, providing insights to operators regarding load conditions during extreme events. These findings highlight the potential of AI-based ensemble forecasting to enhance risk-informed decision making for power system operations.
Load Forecasting for Planning Timescales: Guidance on Large Load Types
Demands on the power system may be widespread across a service area or concentrated at a single location or specific end use. Single large point loads present their own forecasting and planning challenges because of the scale of demand associated with at a specific time, location, or process. Understanding the impacts and characteristics of large point loads can be important to consider in load forecasts and long-term energy system planning when considering system reliability, cost, and resource needs.
Load Forecasting for Planning Timescales: Guidance on Weather and Climate Effects
As the electric power sector faces increasing demands from electrification, data center growth, and climate change, accurate long-term load forecasting has become essential for capacity expansion, transmission planning, and resource adequacy assessments. This presentation outlines the critical role of environmental data—particularly temperature and solar irradiance—in modeling hourly load variability. It emphasizes the importance of integrating historical observations, reanalysis datasets, and climate model projections to capture both natural variability and long-term climate trends. Various data sources and methodologies are reviewed, including statistical adjustments, dynamical modeling, and hybrid approaches such as Quantile Delta Mapping. The limitations of using Typical Meteorological Years (TMYs) are discussed, advocating instead for multi-year weather representations to better capture extreme and compound events. The presentation concludes by recommending a risk-aware, data-diverse approach tailored to the specific analytical context, ensuring more robust and actionable forecasts for future power system planning.
Forecast Trials using the Forecast Arbiter: A Utility Case Study in the Southeast United States
This report presents a case study of a solar forecasting trial conducted by EPRI using the Solar Forecast Arbiter, an open-source platform designed for objective evaluation of short-term solar, wind and load forecasts. The trial involved nine commercial forecast providers submitting forecasts on a recurring basis for four geographically diverse PV sites over a three-month period. The Forecast Arbiter enabled secure data exchange, automated evaluation workflows, and anonymized vendor comparisons, supporting both error-based metrics and relative performance assessments against reference forecasts. Results highlighted significant variability in forecast accuracy across providers, sites, and horizons, with some vendors outperforming baseline models by substantial margins.
Operational Forecasting Framework for Large Electrical Loads
Growing electricity demand from large, high-impact loads such as data centers and EV charging hubs is creating new forecasting challenges for grid operators because these loads behave very differently from traditional demand, with sudden, sharp, and time-specific spikes. Conventional statistical methods struggle to capture these patterns, making accurate short-term forecasting critical for grid reliability, efficient resource use, and better planning. To address this gap, the text introduces a modular, scalable machine-learning framework designed for day-ahead forecasting of large loads, using walk-forward retraining to adapt to evolving behavior. Demonstrated with a public dataset using data centers as a case study, the framework is flexible and broadly applicable to other emerging large-load types beyond data centers.
Accounting for Electrification and Weather in Time-Series Load Forecasts
This report investigates advanced methods for weather normalization and the allocation of system-level forecast information to enhance distribution-level load forecasting for planning applications. Using a real-world utility case study, the report evaluates multiple weather normalization techniques for generating 8,760-hour feeder-level forecasts and comparing their performance. Applying the same case study, the report explores approaches for allocating system-level electrification forecast information to the feeder level and blending them with local distribution-level forecast information over the forecast time horizon.
Full List of Deliverables
- AI Forecasting for Operations During Extreme Events
- Long-Term Electric System Planning Under Load Uncertainty
- Load Forecasting for Distribution Planning: Review of Forecasting Tools
- Load Forecasting for Planning Timescales: Guidance on Weather and Climate Effects
- Load Forecasting for Planning Timescales: Guidance on Large Load Types
- Operational Forecasting Framework for Large Electrical Loads
- Forecast Trials using the Forecast Arbiter: A Utility Case Study in the Southeast United States
- Ensemble Load Forecasting for Operational Timescales
- Quick Insights: Artificial Intelligence Weather Prediction Models
- Operations Forecasting with Wildfire Smoke: Trends and Resources
- Accounting for Electrification and Weather in Time-Series Load Forecasts
- Load Forecasting for Operation Timescales: Data Augmentation and Extreme Events
- Diagnosing Changes in Zonal Load Response to Temperature and Irradiance
- Aligning Transmission and Distribution Forecasts
- Utility Experiences and Trends Regarding Data Centers
- Load Forecasting for Planning Timescales: 2024 Survey
- Quick Insights: Rapid Refresh Forecast System (RRFS)